Skip to content

GitLab

Commit ac98415b authored by WenmuZhou's avatar WenmuZhou
Browse files

merge dygraph

parents af34d785 29929ac6
Show whitespace changes
Inline Side-by-side
Showing with 1437 additions and 97 deletions
ppocr/losses/combined_loss.py
View file @ ac98415b
......@@ -49,11 +49,15 @@ class CombinedLoss(nn.Layer):
loss = loss_func(input, batch, **kargs)
if isinstance(loss, paddle.Tensor):
loss = {"loss_{}_{}".format(str(loss), idx): loss}
weight = self.loss_weight[idx]
for key in loss.keys():
if key == "loss":
loss_all += loss[key] * weight
loss = {key: loss[key] * weight for key in loss}
if "loss" in loss:
loss_all += loss["loss"]
else:
loss_dict["{}_{}".format(key, idx)] = loss[key]
loss_all += paddle.add_n(list(loss.values()))
loss_dict.update(loss)
loss_dict["loss"] = loss_all
return loss_dict
ppocr/losses/distillation_loss.py
View file @ ac98415b
......@@ -44,10 +44,11 @@ class DistillationDMLLoss(DMLLoss):
def __init__(self,
model_name_pairs=[],
act=None,
use_log=False,
key=None,
maps_name=None,
name="dml"):
super().__init__(act=act)
super().__init__(act=act, use_log=use_log)
assert isinstance(model_name_pairs, list)
self.key = key
self.model_name_pairs = self._check_model_name_pairs(model_name_pairs)
......@@ -57,7 +58,8 @@ class DistillationDMLLoss(DMLLoss):
def _check_model_name_pairs(self, model_name_pairs):
if not isinstance(model_name_pairs, list):
return []
elif isinstance(model_name_pairs[0], list) and isinstance(model_name_pairs[0][0], str):
elif isinstance(model_name_pairs[0], list) and isinstance(
model_name_pairs[0][0], str):
return model_name_pairs
else:
return [model_name_pairs]
......@@ -112,8 +114,8 @@ class DistillationDMLLoss(DMLLoss):
loss_dict["{}_{}_{}_{}_{}".format(key, pair[
0], pair[1], map_name, idx)] = loss[key]
else:
loss_dict["{}_{}_{}".format(self.name, self.maps_name[_c],
idx)] = loss
loss_dict["{}_{}_{}".format(self.name, self.maps_name[
_c], idx)] = loss
loss_dict = _sum_loss(loss_dict)
......
ppocr/losses/rec_sar_loss.py 0 → 100644
View file @ ac98415b
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import paddle
from paddle import nn
class SARLoss(nn.Layer):
def __init__(self, **kwargs):
super(SARLoss, self).__init__()
self.loss_func = paddle.nn.loss.CrossEntropyLoss(reduction="mean", ignore_index=96)
def forward(self, predicts, batch):
predict = predicts[:, :-1, :] # ignore last index of outputs to be in same seq_len with targets
label = batch[1].astype("int64")[:, 1:] # ignore first index of target in loss calculation
batch_size, num_steps, num_classes = predict.shape[0], predict.shape[
1], predict.shape[2]
assert len(label.shape) == len(list(predict.shape)) - 1, \
"The target's shape and inputs's shape is [N, d] and [N, num_steps]"
inputs = paddle.reshape(predict, [-1, num_classes])
targets = paddle.reshape(label, [-1])
loss = self.loss_func(inputs, targets)
return {'loss': loss}
ppocr/modeling/backbones/__init__.py
View file @ ac98415b
......@@ -27,8 +27,9 @@ def build_backbone(config, model_type):
from .rec_resnet_fpn import ResNetFPN
from .rec_mv1_enhance import MobileNetV1Enhance
from .rec_nrtr_mtb import MTB
from .rec_resnet_31 import ResNet31
support_dict = [
'MobileNetV1Enhance', 'MobileNetV3', 'ResNet', 'ResNetFPN', 'MTB'
'MobileNetV1Enhance', 'MobileNetV3', 'ResNet', 'ResNetFPN', 'MTB', "ResNet31"
]
elif model_type == "e2e":
from .e2e_resnet_vd_pg import ResNet
......
ppocr/modeling/backbones/rec_resnet_31.py 0 → 100644
View file @ ac98415b
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import paddle
from paddle import ParamAttr
import paddle.nn as nn
import paddle.nn.functional as F
import numpy as np
__all__ = ["ResNet31"]
def conv3x3(in_channel, out_channel, stride=1):
return nn.Conv2D(
in_channel,
out_channel,
kernel_size=3,
stride=stride,
padding=1,
bias_attr=False
)
class BasicBlock(nn.Layer):
expansion = 1
def __init__(self, in_channels, channels, stride=1, downsample=False):
super().__init__()
self.conv1 = conv3x3(in_channels, channels, stride)
self.bn1 = nn.BatchNorm2D(channels)
self.relu = nn.ReLU()
self.conv2 = conv3x3(channels, channels)
self.bn2 = nn.BatchNorm2D(channels)
self.downsample = downsample
if downsample:
self.downsample = nn.Sequential(
nn.Conv2D(in_channels, channels * self.expansion, 1, stride, bias_attr=False),
nn.BatchNorm2D(channels * self.expansion),
)
else:
self.downsample = nn.Sequential()
self.stride = stride
def forward(self, x):
residual = x
out = self.conv1(x)
out = self.bn1(out)
out = self.relu(out)
out = self.conv2(out)
out = self.bn2(out)
if self.downsample:
residual = self.downsample(x)
out += residual
out = self.relu(out)
return out
class ResNet31(nn.Layer):
'''
Args:
in_channels (int): Number of channels of input image tensor.
layers (list[int]): List of BasicBlock number for each stage.
channels (list[int]): List of out_channels of Conv2d layer.
out_indices (None | Sequence[int]): Indices of output stages.
last_stage_pool (bool): If True, add `MaxPool2d` layer to last stage.
'''
def __init__(self,
in_channels=3,
layers=[1, 2, 5, 3],
channels=[64, 128, 256, 256, 512, 512, 512],
out_indices=None,
last_stage_pool=False):
super(ResNet31, self).__init__()
assert isinstance(in_channels, int)
assert isinstance(last_stage_pool, bool)
self.out_indices = out_indices
self.last_stage_pool = last_stage_pool
# conv 1 (Conv Conv)
self.conv1_1 = nn.Conv2D(in_channels, channels[0], kernel_size=3, stride=1, padding=1)
self.bn1_1 = nn.BatchNorm2D(channels[0])
self.relu1_1 = nn.ReLU()
self.conv1_2 = nn.Conv2D(channels[0], channels[1], kernel_size=3, stride=1, padding=1)
self.bn1_2 = nn.BatchNorm2D(channels[1])
self.relu1_2 = nn.ReLU()
# conv 2 (Max-pooling, Residual block, Conv)
self.pool2 = nn.MaxPool2D(kernel_size=2, stride=2, padding=0, ceil_mode=True)
self.block2 = self._make_layer(channels[1], channels[2], layers[0])
self.conv2 = nn.Conv2D(channels[2], channels[2], kernel_size=3, stride=1, padding=1)
self.bn2 = nn.BatchNorm2D(channels[2])
self.relu2 = nn.ReLU()
# conv 3 (Max-pooling, Residual block, Conv)
self.pool3 = nn.MaxPool2D(kernel_size=2, stride=2, padding=0, ceil_mode=True)
self.block3 = self._make_layer(channels[2], channels[3], layers[1])
self.conv3 = nn.Conv2D(channels[3], channels[3], kernel_size=3, stride=1, padding=1)
self.bn3 = nn.BatchNorm2D(channels[3])
self.relu3 = nn.ReLU()
# conv 4 (Max-pooling, Residual block, Conv)
self.pool4 = nn.MaxPool2D(kernel_size=(2, 1), stride=(2, 1), padding=0, ceil_mode=True)
self.block4 = self._make_layer(channels[3], channels[4], layers[2])
self.conv4 = nn.Conv2D(channels[4], channels[4], kernel_size=3, stride=1, padding=1)
self.bn4 = nn.BatchNorm2D(channels[4])
self.relu4 = nn.ReLU()
# conv 5 ((Max-pooling), Residual block, Conv)
self.pool5 = None
if self.last_stage_pool:
self.pool5 = nn.MaxPool2D(kernel_size=2, stride=2, padding=0, ceil_mode=True)
self.block5 = self._make_layer(channels[4], channels[5], layers[3])
self.conv5 = nn.Conv2D(channels[5], channels[5], kernel_size=3, stride=1, padding=1)
self.bn5 = nn.BatchNorm2D(channels[5])
self.relu5 = nn.ReLU()
self.out_channels = channels[-1]
def _make_layer(self, input_channels, output_channels, blocks):
layers = []
for _ in range(blocks):
downsample = None
if input_channels != output_channels:
downsample = nn.Sequential(
nn.Conv2D(
input_channels,
output_channels,
kernel_size=1,
stride=1,
bias_attr=False),
nn.BatchNorm2D(output_channels),
)
layers.append(BasicBlock(input_channels, output_channels, downsample=downsample))
input_channels = output_channels
return nn.Sequential(*layers)
def forward(self, x):
x = self.conv1_1(x)
x = self.bn1_1(x)
x = self.relu1_1(x)
x = self.conv1_2(x)
x = self.bn1_2(x)
x = self.relu1_2(x)
outs = []
for i in range(4):
layer_index = i + 2
pool_layer = getattr(self, f'pool{layer_index}')
block_layer = getattr(self, f'block{layer_index}')
conv_layer = getattr(self, f'conv{layer_index}')
bn_layer = getattr(self, f'bn{layer_index}')
relu_layer = getattr(self, f'relu{layer_index}')
if pool_layer is not None:
x = pool_layer(x)
x = block_layer(x)
x = conv_layer(x)
x = bn_layer(x)
x= relu_layer(x)
outs.append(x)
if self.out_indices is not None:
return tuple([outs[i] for i in self.out_indices])
return x
ppocr/modeling/heads/__init__.py
View file @ ac98415b
......@@ -28,13 +28,14 @@ def build_head(config):
from .rec_att_head import AttentionHead
from .rec_srn_head import SRNHead
from .rec_nrtr_head import Transformer
from .rec_sar_head import SARHead
# cls head
from .cls_head import ClsHead
support_dict = [
'DBHead', 'PSEHead', 'EASTHead', 'SASTHead', 'CTCHead', 'ClsHead',
'AttentionHead', 'SRNHead', 'PGHead', 'Transformer',
'TableAttentionHead'
'TableAttentionHead', 'SARHead'
]
#table head
......
ppocr/modeling/heads/rec_sar_head.py 0 → 100644
View file @ ac98415b
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import math
import paddle
from paddle import ParamAttr
import paddle.nn as nn
import paddle.nn.functional as F
class SAREncoder(nn.Layer):
"""
Args:
enc_bi_rnn (bool): If True, use bidirectional RNN in encoder.
enc_drop_rnn (float): Dropout probability of RNN layer in encoder.
enc_gru (bool): If True, use GRU, else LSTM in encoder.
d_model (int): Dim of channels from backbone.
d_enc (int): Dim of encoder RNN layer.
mask (bool): If True, mask padding in RNN sequence.
"""
def __init__(self,
enc_bi_rnn=False,
enc_drop_rnn=0.1,
enc_gru=False,
d_model=512,
d_enc=512,
mask=True,
**kwargs):
super().__init__()
assert isinstance(enc_bi_rnn, bool)
assert isinstance(enc_drop_rnn, (int, float))
assert 0 <= enc_drop_rnn < 1.0
assert isinstance(enc_gru, bool)
assert isinstance(d_model, int)
assert isinstance(d_enc, int)
assert isinstance(mask, bool)
self.enc_bi_rnn = enc_bi_rnn
self.enc_drop_rnn = enc_drop_rnn
self.mask = mask
# LSTM Encoder
if enc_bi_rnn:
direction = 'bidirectional'
else:
direction = 'forward'
kwargs = dict(
input_size=d_model,
hidden_size=d_enc,
num_layers=2,
time_major=False,
dropout=enc_drop_rnn,
direction=direction)
if enc_gru:
self.rnn_encoder = nn.GRU(**kwargs)
else:
self.rnn_encoder = nn.LSTM(**kwargs)
# global feature transformation
encoder_rnn_out_size = d_enc * (int(enc_bi_rnn) + 1)
self.linear = nn.Linear(encoder_rnn_out_size, encoder_rnn_out_size)
def forward(self, feat, img_metas=None):
if img_metas is not None:
assert len(img_metas[0]) == feat.shape[0]
valid_ratios = None
if img_metas is not None and self.mask:
valid_ratios = img_metas[-1]
h_feat = feat.shape[2] # bsz c h w
feat_v = F.max_pool2d(
feat, kernel_size=(h_feat, 1), stride=1, padding=0)
feat_v = feat_v.squeeze(2) # bsz * C * W
feat_v = paddle.transpose(feat_v, perm=[0, 2, 1]) # bsz * W * C
holistic_feat = self.rnn_encoder(feat_v)[0] # bsz * T * C
if valid_ratios is not None:
valid_hf = []
T = holistic_feat.shape[1]
for i, valid_ratio in enumerate(valid_ratios):
valid_step = min(T, math.ceil(T * valid_ratio)) - 1
valid_hf.append(holistic_feat[i, valid_step, :])
valid_hf = paddle.stack(valid_hf, axis=0)
else:
valid_hf = holistic_feat[:, -1, :] # bsz * C
holistic_feat = self.linear(valid_hf) # bsz * C
return holistic_feat
class BaseDecoder(nn.Layer):
def __init__(self, **kwargs):
super().__init__()
def forward_train(self, feat, out_enc, targets, img_metas):
raise NotImplementedError
def forward_test(self, feat, out_enc, img_metas):
raise NotImplementedError
def forward(self,
feat,
out_enc,
label=None,
img_metas=None,
train_mode=True):
self.train_mode = train_mode
if train_mode:
return self.forward_train(feat, out_enc, label, img_metas)
return self.forward_test(feat, out_enc, img_metas)
class ParallelSARDecoder(BaseDecoder):
"""
Args:
out_channels (int): Output class number.
enc_bi_rnn (bool): If True, use bidirectional RNN in encoder.
dec_bi_rnn (bool): If True, use bidirectional RNN in decoder.
dec_drop_rnn (float): Dropout of RNN layer in decoder.
dec_gru (bool): If True, use GRU, else LSTM in decoder.
d_model (int): Dim of channels from backbone.
d_enc (int): Dim of encoder RNN layer.
d_k (int): Dim of channels of attention module.
pred_dropout (float): Dropout probability of prediction layer.
max_seq_len (int): Maximum sequence length for decoding.
mask (bool): If True, mask padding in feature map.
start_idx (int): Index of start token.
padding_idx (int): Index of padding token.
pred_concat (bool): If True, concat glimpse feature from
attention with holistic feature and hidden state.
"""
def __init__(
self,
out_channels, # 90 + unknown + start + padding
enc_bi_rnn=False,
dec_bi_rnn=False,
dec_drop_rnn=0.0,
dec_gru=False,
d_model=512,
d_enc=512,
d_k=64,
pred_dropout=0.1,
max_text_length=30,
mask=True,
pred_concat=True,
**kwargs):
super().__init__()
self.num_classes = out_channels
self.enc_bi_rnn = enc_bi_rnn
self.d_k = d_k
self.start_idx = out_channels - 2
self.padding_idx = out_channels - 1
self.max_seq_len = max_text_length
self.mask = mask
self.pred_concat = pred_concat
encoder_rnn_out_size = d_enc * (int(enc_bi_rnn) + 1)
decoder_rnn_out_size = encoder_rnn_out_size * (int(dec_bi_rnn) + 1)
# 2D attention layer
self.conv1x1_1 = nn.Linear(decoder_rnn_out_size, d_k)
self.conv3x3_1 = nn.Conv2D(
d_model, d_k, kernel_size=3, stride=1, padding=1)
self.conv1x1_2 = nn.Linear(d_k, 1)
# Decoder RNN layer
if dec_bi_rnn:
direction = 'bidirectional'
else:
direction = 'forward'
kwargs = dict(
input_size=encoder_rnn_out_size,
hidden_size=encoder_rnn_out_size,
num_layers=2,
time_major=False,
dropout=dec_drop_rnn,
direction=direction)
if dec_gru:
self.rnn_decoder = nn.GRU(**kwargs)
else:
self.rnn_decoder = nn.LSTM(**kwargs)
# Decoder input embedding
self.embedding = nn.Embedding(
self.num_classes,
encoder_rnn_out_size,
padding_idx=self.padding_idx)
# Prediction layer
self.pred_dropout = nn.Dropout(pred_dropout)
pred_num_classes = self.num_classes - 1
if pred_concat:
fc_in_channel = decoder_rnn_out_size + d_model + d_enc
else:
fc_in_channel = d_model
self.prediction = nn.Linear(fc_in_channel, pred_num_classes)
def _2d_attention(self,
decoder_input,
feat,
holistic_feat,
valid_ratios=None):
y = self.rnn_decoder(decoder_input)[0]
# y: bsz * (seq_len + 1) * hidden_size
attn_query = self.conv1x1_1(y) # bsz * (seq_len + 1) * attn_size
bsz, seq_len, attn_size = attn_query.shape
attn_query = paddle.unsqueeze(attn_query, axis=[3, 4])
# (bsz, seq_len + 1, attn_size, 1, 1)
attn_key = self.conv3x3_1(feat)
# bsz * attn_size * h * w
attn_key = attn_key.unsqueeze(1)
# bsz * 1 * attn_size * h * w
attn_weight = paddle.tanh(paddle.add(attn_key, attn_query))
# bsz * (seq_len + 1) * attn_size * h * w
attn_weight = paddle.transpose(attn_weight, perm=[0, 1, 3, 4, 2])
# bsz * (seq_len + 1) * h * w * attn_size
attn_weight = self.conv1x1_2(attn_weight)
# bsz * (seq_len + 1) * h * w * 1
bsz, T, h, w, c = attn_weight.shape
assert c == 1
if valid_ratios is not None:
# cal mask of attention weight
for i, valid_ratio in enumerate(valid_ratios):
valid_width = min(w, math.ceil(w * valid_ratio))
attn_weight[i, :, :, valid_width:, :] = float('-inf')
attn_weight = paddle.reshape(attn_weight, [bsz, T, -1])
attn_weight = F.softmax(attn_weight, axis=-1)
attn_weight = paddle.reshape(attn_weight, [bsz, T, h, w, c])
attn_weight = paddle.transpose(attn_weight, perm=[0, 1, 4, 2, 3])
# attn_weight: bsz * T * c * h * w
# feat: bsz * c * h * w
attn_feat = paddle.sum(paddle.multiply(feat.unsqueeze(1), attn_weight),
(3, 4),
keepdim=False)
# bsz * (seq_len + 1) * C
# Linear transformation
if self.pred_concat:
hf_c = holistic_feat.shape[-1]
holistic_feat = paddle.expand(
holistic_feat, shape=[bsz, seq_len, hf_c])
y = self.prediction(paddle.concat((y, attn_feat, holistic_feat), 2))
else:
y = self.prediction(attn_feat)
# bsz * (seq_len + 1) * num_classes
if self.train_mode:
y = self.pred_dropout(y)
return y
def forward_train(self, feat, out_enc, label, img_metas):
'''
img_metas: [label, valid_ratio]
'''
if img_metas is not None:
assert len(img_metas[0]) == feat.shape[0]
valid_ratios = None
if img_metas is not None and self.mask:
valid_ratios = img_metas[-1]
label = label.cuda()
lab_embedding = self.embedding(label)
# bsz * seq_len * emb_dim
out_enc = out_enc.unsqueeze(1)
# bsz * 1 * emb_dim
in_dec = paddle.concat((out_enc, lab_embedding), axis=1)
# bsz * (seq_len + 1) * C
out_dec = self._2d_attention(
in_dec, feat, out_enc, valid_ratios=valid_ratios)
# bsz * (seq_len + 1) * num_classes
return out_dec[:, 1:, :] # bsz * seq_len * num_classes
def forward_test(self, feat, out_enc, img_metas):
if img_metas is not None:
assert len(img_metas[0]) == feat.shape[0]
valid_ratios = None
if img_metas is not None and self.mask:
valid_ratios = img_metas[-1]
seq_len = self.max_seq_len
bsz = feat.shape[0]
start_token = paddle.full(
(bsz, ), fill_value=self.start_idx, dtype='int64')
# bsz
start_token = self.embedding(start_token)
# bsz * emb_dim
emb_dim = start_token.shape[1]
start_token = start_token.unsqueeze(1)
start_token = paddle.expand(start_token, shape=[bsz, seq_len, emb_dim])
# bsz * seq_len * emb_dim
out_enc = out_enc.unsqueeze(1)
# bsz * 1 * emb_dim
decoder_input = paddle.concat((out_enc, start_token), axis=1)
# bsz * (seq_len + 1) * emb_dim
outputs = []
for i in range(1, seq_len + 1):
decoder_output = self._2d_attention(
decoder_input, feat, out_enc, valid_ratios=valid_ratios)
char_output = decoder_output[:, i, :] # bsz * num_classes
char_output = F.softmax(char_output, -1)
outputs.append(char_output)
max_idx = paddle.argmax(char_output, axis=1, keepdim=False)
char_embedding = self.embedding(max_idx) # bsz * emb_dim
if i < seq_len:
decoder_input[:, i + 1, :] = char_embedding
outputs = paddle.stack(outputs, 1) # bsz * seq_len * num_classes
return outputs
class SARHead(nn.Layer):
def __init__(self,
out_channels,
enc_bi_rnn=False,
enc_drop_rnn=0.1,
enc_gru=False,
dec_bi_rnn=False,
dec_drop_rnn=0.0,
dec_gru=False,
d_k=512,
pred_dropout=0.1,
max_text_length=30,
pred_concat=True,
**kwargs):
super(SARHead, self).__init__()
# encoder module
self.encoder = SAREncoder(
enc_bi_rnn=enc_bi_rnn, enc_drop_rnn=enc_drop_rnn, enc_gru=enc_gru)
# decoder module
self.decoder = ParallelSARDecoder(
out_channels=out_channels,
enc_bi_rnn=enc_bi_rnn,
dec_bi_rnn=dec_bi_rnn,
dec_drop_rnn=dec_drop_rnn,
dec_gru=dec_gru,
d_k=d_k,
pred_dropout=pred_dropout,
max_text_length=max_text_length,
pred_concat=pred_concat)
def forward(self, feat, targets=None):
'''
img_metas: [label, valid_ratio]
'''
holistic_feat = self.encoder(feat, targets) # bsz c
if self.training:
label = targets[0] # label
label = paddle.to_tensor(label, dtype='int64')
final_out = self.decoder(
feat, holistic_feat, label, img_metas=targets)
if not self.training:
final_out = self.decoder(
feat,
holistic_feat,
label=None,
img_metas=targets,
train_mode=False)
# (bsz, seq_len, num_classes)
return final_out
ppocr/postprocess/__init__.py
View file @ ac98415b
......@@ -25,7 +25,7 @@ from .db_postprocess import DBPostProcess, DistillationDBPostProcess
from .east_postprocess import EASTPostProcess
from .sast_postprocess import SASTPostProcess
from .rec_postprocess import CTCLabelDecode, AttnLabelDecode, SRNLabelDecode, DistillationCTCLabelDecode, NRTRLabelDecode, \
TableLabelDecode
TableLabelDecode, SARLabelDecode
from .cls_postprocess import ClsPostProcess
from .pg_postprocess import PGPostProcess
from .pse_postprocess import PSEPostProcess
......@@ -35,8 +35,8 @@ def build_post_process(config, global_config=None):
support_dict = [
'DBPostProcess', 'PSEPostProcess', 'EASTPostProcess', 'SASTPostProcess',
'CTCLabelDecode', 'AttnLabelDecode', 'ClsPostProcess', 'SRNLabelDecode',
'PGPostProcess', 'DistillationCTCLabelDecode', 'NRTRLabelDecode',
'TableLabelDecode', 'DistillationDBPostProcess'
'PGPostProcess', 'DistillationCTCLabelDecode', 'TableLabelDecode',
'DistillationDBPostProcess', 'NRTRLabelDecode', 'SARLabelDecode'
]
config = copy.deepcopy(config)
......
ppocr/postprocess/rec_postprocess.py
View file @ ac98415b
......@@ -15,6 +15,7 @@ import numpy as np
import string
import paddle
from paddle.nn import functional as F
import re
class BaseRecLabelDecode(object):
......@@ -171,15 +172,15 @@ class NRTRLabelDecode(BaseRecLabelDecode):
if preds.dtype == paddle.int64:
if isinstance(preds, paddle.Tensor):
preds = preds.numpy()
if preds[0][0]==2:
preds_idx = preds[:,1:]
if preds[0][0] == 2:
preds_idx = preds[:, 1:]
else:
preds_idx = preds
text = self.decode(preds_idx)
if label is None:
return text
label = self.decode(label[:,1:])
label = self.decode(label[:, 1:])
else:
if isinstance(preds, paddle.Tensor):
preds = preds.numpy()
......@@ -188,11 +189,11 @@ class NRTRLabelDecode(BaseRecLabelDecode):
text = self.decode(preds_idx, preds_prob, is_remove_duplicate=False)
if label is None:
return text
label = self.decode(label[:,1:])
label = self.decode(label[:, 1:])
return text, label
def add_special_char(self, dict_character):
dict_character = ['blank','<unk>','<s>','</s>'] + dict_character
dict_character = ['blank', '<unk>', '<s>', '</s>'] + dict_character
return dict_character
def decode(self, text_index, text_prob=None, is_remove_duplicate=False):
......@@ -206,7 +207,8 @@ class NRTRLabelDecode(BaseRecLabelDecode):
if text_index[batch_idx][idx] == 3: # end
break
try:
char_list.append(self.character[int(text_index[batch_idx][idx])])
char_list.append(self.character[int(text_index[batch_idx][
idx])])
except:
continue
if text_prob is not None:
......@@ -218,7 +220,6 @@ class NRTRLabelDecode(BaseRecLabelDecode):
return result_list
class AttnLabelDecode(BaseRecLabelDecode):
""" Convert between text-label and text-index """
......@@ -256,7 +257,8 @@ class AttnLabelDecode(BaseRecLabelDecode):
if idx > 0 and text_index[batch_idx][idx - 1] == text_index[
batch_idx][idx]:
continue
char_list.append(self.character[int(text_index[batch_idx][idx])])
char_list.append(self.character[int(text_index[batch_idx][
idx])])
if text_prob is not None:
conf_list.append(text_prob[batch_idx][idx])
else:
......@@ -386,10 +388,9 @@ class SRNLabelDecode(BaseRecLabelDecode):
class TableLabelDecode(object):
""" """
def __init__(self,
character_dict_path,
**kwargs):
list_character, list_elem = self.load_char_elem_dict(character_dict_path)
def __init__(self, character_dict_path, **kwargs):
list_character, list_elem = self.load_char_elem_dict(
character_dict_path)
list_character = self.add_special_char(list_character)
list_elem = self.add_special_char(list_elem)
self.dict_character = {}
......@@ -408,7 +409,8 @@ class TableLabelDecode(object):
list_elem = []
with open(character_dict_path, "rb") as fin:
lines = fin.readlines()
substr = lines[0].decode('utf-8').strip("\n").strip("\r\n").split("\t")
substr = lines[0].decode('utf-8').strip("\n").strip("\r\n").split(
"\t")
character_num = int(substr[0])
elem_num = int(substr[1])
for cno in range(1, 1 + character_num):
......@@ -428,14 +430,14 @@ class TableLabelDecode(object):
def __call__(self, preds):
structure_probs = preds['structure_probs']
loc_preds = preds['loc_preds']
if isinstance(structure_probs,paddle.Tensor):
if isinstance(structure_probs, paddle.Tensor):
structure_probs = structure_probs.numpy()
if isinstance(loc_preds,paddle.Tensor):
if isinstance(loc_preds, paddle.Tensor):
loc_preds = loc_preds.numpy()
structure_idx = structure_probs.argmax(axis=2)
structure_probs = structure_probs.max(axis=2)
structure_str, structure_pos, result_score_list, result_elem_idx_list = self.decode(structure_idx,
structure_probs, 'elem')
structure_str, structure_pos, result_score_list, result_elem_idx_list = self.decode(
structure_idx, structure_probs, 'elem')
res_html_code_list = []
res_loc_list = []
batch_num = len(structure_str)
......@@ -450,8 +452,13 @@ class TableLabelDecode(object):
res_loc = np.array(res_loc)
res_html_code_list.append(res_html_code)
res_loc_list.append(res_loc)
return {'res_html_code': res_html_code_list, 'res_loc': res_loc_list, 'res_score_list': result_score_list,
'res_elem_idx_list': result_elem_idx_list,'structure_str_list':structure_str}
return {
'res_html_code': res_html_code_list,
'res_loc': res_loc_list,
'res_score_list': result_score_list,
'res_elem_idx_list': result_elem_idx_list,
'structure_str_list': structure_str
}
def decode(self, text_index, structure_probs, char_or_elem):
"""convert text-label into text-index.
......@@ -516,3 +523,82 @@ class TableLabelDecode(object):
assert False, "Unsupport type %s in char_or_elem" \
% char_or_elem
return idx
class SARLabelDecode(BaseRecLabelDecode):
""" Convert between text-label and text-index """
def __init__(self,
character_dict_path=None,
character_type='ch',
use_space_char=False,
**kwargs):
super(SARLabelDecode, self).__init__(character_dict_path,
character_type, use_space_char)
self.rm_symbol = kwargs.get('rm_symbol', False)
def add_special_char(self, dict_character):
beg_end_str = "<BOS/EOS>"
unknown_str = "<UKN>"
padding_str = "<PAD>"
dict_character = dict_character + [unknown_str]
self.unknown_idx = len(dict_character) - 1
dict_character = dict_character + [beg_end_str]
self.start_idx = len(dict_character) - 1
self.end_idx = len(dict_character) - 1
dict_character = dict_character + [padding_str]
self.padding_idx = len(dict_character) - 1
return dict_character
def decode(self, text_index, text_prob=None, is_remove_duplicate=False):
""" convert text-index into text-label. """
result_list = []
ignored_tokens = self.get_ignored_tokens()
batch_size = len(text_index)
for batch_idx in range(batch_size):
char_list = []
conf_list = []
for idx in range(len(text_index[batch_idx])):
if text_index[batch_idx][idx] in ignored_tokens:
continue
if int(text_index[batch_idx][idx]) == int(self.end_idx):
if text_prob is None and idx == 0:
continue
else:
break
if is_remove_duplicate:
# only for predict
if idx > 0 and text_index[batch_idx][idx - 1] == text_index[
batch_idx][idx]:
continue
char_list.append(self.character[int(text_index[batch_idx][
idx])])
if text_prob is not None:
conf_list.append(text_prob[batch_idx][idx])
else:
conf_list.append(1)
text = ''.join(char_list)
if self.rm_symbol:
comp = re.compile('[^A-Z^a-z^0-9^\u4e00-\u9fa5]')
text = text.lower()
text = comp.sub('', text)
result_list.append((text, np.mean(conf_list)))
return result_list
def __call__(self, preds, label=None, *args, **kwargs):
if isinstance(preds, paddle.Tensor):
preds = preds.numpy()
preds_idx = preds.argmax(axis=2)
preds_prob = preds.max(axis=2)
text = self.decode(preds_idx, preds_prob, is_remove_duplicate=False)
if label is None:
return text
label = self.decode(label, is_remove_duplicate=False)
return text, label
def get_ignored_tokens(self):
return [self.padding_idx]
ppocr/utils/save_load.py
View file @ ac98415b
......@@ -116,6 +116,7 @@ def load_dygraph_params(config, model, logger, optimizer):
logger.info(f"loaded pretrained_model successful from {pm}")
return {}
def load_pretrained_params(model, path):
if path is None:
return False
......@@ -138,6 +139,7 @@ def load_pretrained_params(model, path):
print(f"load pretrain successful from {path}")
return model
def save_model(model,
optimizer,
model_path,
......
tests/configs/det_mv3_db.yml 0 → 100644
View file @ ac98415b
Global:
use_gpu: false
epoch_num: 5
log_smooth_window: 20
print_batch_step: 1
save_model_dir: ./output/db_mv3/
save_epoch_step: 1200
# evaluation is run every 2000 iterations
eval_batch_step: [0, 400]
cal_metric_during_train: False
pretrained_model: ./pretrain_models/MobileNetV3_large_x0_5_pretrained
checkpoints:
save_inference_dir:
use_visualdl: False
infer_img: doc/imgs_en/img_10.jpg
save_res_path: ./output/det_db/predicts_db.txt
Architecture:
model_type: det
algorithm: DB
Transform:
Backbone:
name: MobileNetV3
scale: 0.5
model_name: large
disable_se: True
Neck:
name: DBFPN
out_channels: 96
Head:
name: DBHead
k: 50
Loss:
name: DBLoss
balance_loss: true
main_loss_type: DiceLoss
alpha: 5
beta: 10
ohem_ratio: 3
Optimizer:
name: Adam #Momentum
#momentum: 0.9
beta1: 0.9
beta2: 0.999
lr:
learning_rate: 0.001
regularizer:
name: 'L2'
factor: 0
PostProcess:
name: DBPostProcess
thresh: 0.3
box_thresh: 0.6
max_candidates: 1000
unclip_ratio: 1.5
Metric:
name: DetMetric
main_indicator: hmean
Train:
dataset:
name: SimpleDataSet
data_dir: ./train_data/icdar2015/text_localization/
label_file_list:
- ./train_data/icdar2015/text_localization/train_icdar2015_label.txt
ratio_list: [1.0]
transforms:
- DecodeImage: # load image
img_mode: BGR
channel_first: False
- DetLabelEncode: # Class handling label
- Resize:
# size: [640, 640]
- MakeBorderMap:
shrink_ratio: 0.4
thresh_min: 0.3
thresh_max: 0.7
- MakeShrinkMap:
shrink_ratio: 0.4
min_text_size: 8
- NormalizeImage:
scale: 1./255.
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: 'hwc'
- ToCHWImage:
- KeepKeys:
keep_keys: ['image', 'threshold_map', 'threshold_mask', 'shrink_map', 'shrink_mask'] # the order of the dataloader list
loader:
shuffle: False
drop_last: False
batch_size_per_card: 1
num_workers: 0
use_shared_memory: False
Eval:
dataset:
name: SimpleDataSet
data_dir: ./train_data/icdar2015/text_localization/
label_file_list:
- ./train_data/icdar2015/text_localization/test_icdar2015_label.txt
transforms:
- DecodeImage: # load image
img_mode: BGR
channel_first: False
- DetLabelEncode: # Class handling label
- DetResizeForTest:
image_shape: [736, 1280]
- NormalizeImage:
scale: 1./255.
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: 'hwc'
- ToCHWImage:
- KeepKeys:
keep_keys: ['image', 'shape', 'polys', 'ignore_tags']
loader:
shuffle: False
drop_last: False
batch_size_per_card: 1 # must be 1
num_workers: 0
use_shared_memory: False
tests/configs/det_r50_vd_db.yml 0 → 100644
View file @ ac98415b
Global:
use_gpu: false
epoch_num: 5
log_smooth_window: 20
print_batch_step: 1
save_model_dir: ./output/db_mv3/
save_epoch_step: 1200
# evaluation is run every 2000 iterations
eval_batch_step: [0, 400]
cal_metric_during_train: False
pretrained_model: ./pretrain_models/MobileNetV3_large_x0_5_pretrained
checkpoints:
save_inference_dir:
use_visualdl: False
infer_img: doc/imgs_en/img_10.jpg
save_res_path: ./output/det_db/predicts_db.txt
Architecture:
model_type: det
algorithm: DB
Transform:
Backbone:
name: ResNet #MobileNetV3
layers: 50
Neck:
name: DBFPN
out_channels: 256
Head:
name: DBHead
k: 50
Loss:
name: DBLoss
balance_loss: true
main_loss_type: DiceLoss
alpha: 5 #5
beta: 10 #10
ohem_ratio: 3
Optimizer:
name: Adam #Momentum
#momentum: 0.9
beta1: 0.9
beta2: 0.999
lr:
learning_rate: 0.001
regularizer:
name: 'L2'
factor: 0
PostProcess:
name: DBPostProcess
thresh: 0.3
box_thresh: 0.6
max_candidates: 1000
unclip_ratio: 1.5
Metric:
name: DetMetric
main_indicator: hmean
Train:
dataset:
name: SimpleDataSet
data_dir: ./train_data/icdar2015/text_localization/
label_file_list:
- ./train_data/icdar2015/text_localization/train_icdar2015_label.txt
ratio_list: [1.0]
transforms:
- DecodeImage: # load image
img_mode: BGR
channel_first: False
- DetLabelEncode: # Class handling label
- Resize:
# size: [640, 640]
- MakeBorderMap:
shrink_ratio: 0.4
thresh_min: 0.3
thresh_max: 0.7
- MakeShrinkMap:
shrink_ratio: 0.4
min_text_size: 8
- NormalizeImage:
scale: 1./255.
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: 'hwc'
- ToCHWImage:
- KeepKeys:
keep_keys: ['image', 'threshold_map', 'threshold_mask', 'shrink_map', 'shrink_mask'] # the order of the dataloader list
loader:
shuffle: False
drop_last: False
batch_size_per_card: 1
num_workers: 0
use_shared_memory: False
Eval:
dataset:
name: SimpleDataSet
data_dir: ./train_data/icdar2015/text_localization/
label_file_list:
- ./train_data/icdar2015/text_localization/test_icdar2015_label.txt
transforms:
- DecodeImage: # load image
img_mode: BGR
channel_first: False
- DetLabelEncode: # Class handling label
- DetResizeForTest:
image_shape: [736, 1280]
- NormalizeImage:
scale: 1./255.
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: 'hwc'
- ToCHWImage:
- KeepKeys:
keep_keys: ['image', 'shape', 'polys', 'ignore_tags']
loader:
shuffle: False
drop_last: False
batch_size_per_card: 1 # must be 1
num_workers: 0
use_shared_memory: False
tests/configs/rec_icdar15_r34_train.yml 0 → 100644
View file @ ac98415b
Global:
use_gpu: true
epoch_num: 72
log_smooth_window: 20
print_batch_step: 10
save_model_dir: ./output/rec/ic15/
save_epoch_step: 3
# evaluation is run every 2000 iterations
eval_batch_step: [0, 2000]
cal_metric_during_train: True
pretrained_model:
checkpoints:
save_inference_dir: ./
use_visualdl: False
infer_img: doc/imgs_words_en/word_10.png
# for data or label process
character_dict_path: ppocr/utils/en_dict.txt
character_type: EN
max_text_length: 25
infer_mode: False
use_space_char: False
save_res_path: ./output/rec/predicts_ic15.txt
Optimizer:
name: Adam
beta1: 0.9
beta2: 0.999
lr:
learning_rate: 0.0005
regularizer:
name: 'L2'
factor: 0
Architecture:
model_type: rec
algorithm: CRNN
Transform:
Backbone:
name: ResNet
layers: 34
Neck:
name: SequenceEncoder
encoder_type: rnn
hidden_size: 256
Head:
name: CTCHead
fc_decay: 0
Loss:
name: CTCLoss
PostProcess:
name: CTCLabelDecode
Metric:
name: RecMetric
main_indicator: acc
Train:
dataset:
name: SimpleDataSet
data_dir: ./train_data/ic15_data/
label_file_list: ["./train_data/ic15_data/rec_gt_train.txt"]
transforms:
- DecodeImage: # load image
img_mode: BGR
channel_first: False
- CTCLabelEncode: # Class handling label
- RecResizeImg:
image_shape: [3, 32, 100]
- KeepKeys:
keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
loader:
shuffle: True
batch_size_per_card: 256
drop_last: True
num_workers: 8
use_shared_memory: False
Eval:
dataset:
name: SimpleDataSet
data_dir: ./train_data/ic15_data
label_file_list: ["./train_data/ic15_data/rec_gt_test.txt"]
transforms:
- DecodeImage: # load image
img_mode: BGR
channel_first: False
- CTCLabelEncode: # Class handling label
- RecResizeImg:
image_shape: [3, 32, 100]
- KeepKeys:
keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
loader:
shuffle: False
drop_last: False
batch_size_per_card: 256
num_workers: 4
use_shared_memory: False
tests/ocr_det_params.txt
View file @ ac98415b
......@@ -13,34 +13,34 @@ train_infer_img_dir:./train_data/icdar2015/text_localization/ch4_test_images/
null:null
##
trainer:norm_train|pact_train
norm_train:tools/train.py -c configs/det/det_mv3_db.yml -o Global.pretrained_model=./pretrain_models/MobileNetV3_large_x0_5_pretrained
pact_train:deploy/slim/quantization/quant.py -c configs/det/det_mv3_db.yml -o
fpgm_train:deploy/slim/prune/sensitivity_anal.py -c configs/det/det_mv3_db.yml -o Global.pretrained_model=./pretrain_models/det_mv3_db_v2.0_train/best_accuracy
norm_train:tools/train.py -c tests/configs/det_mv3_db.yml -o Global.pretrained_model=./pretrain_models/MobileNetV3_large_x0_5_pretrained
pact_train:deploy/slim/quantization/quant.py -c tests/configs/det_mv3_db.yml -o
fpgm_train:deploy/slim/prune/sensitivity_anal.py -c tests/configs/det_mv3_db.yml -o Global.pretrained_model=./pretrain_models/det_mv3_db_v2.0_train/best_accuracy
distill_train:null
null:null
null:null
##
===========================eval_params===========================
eval:tools/eval.py -c configs/det/det_mv3_db.yml -o
eval:tools/eval.py -c tests/configs/det_mv3_db.yml -o
null:null
##
===========================infer_params===========================
Global.save_inference_dir:./output/
Global.pretrained_model:
norm_export:tools/export_model.py -c configs/det/det_mv3_db.yml -o
quant_export:deploy/slim/quantization/export_model.py -c configs/det/det_mv3_db.yml -o
fpgm_export:deploy/slim/prune/export_prune_model.py -c configs/det/det_mv3_db.yml -o
norm_export:tools/export_model.py -c tests/configs/det_mv3_db.yml -o
quant_export:deploy/slim/quantization/export_model.py -c tests/configs/det_mv3_db.yml -o
fpgm_export:deploy/slim/prune/export_prune_model.py -c tests/configs/det_mv3_db.yml -o
distill_export:null
export1:null
export2:null
##
infer_model:./inference/ch_ppocr_mobile_v2.0_det_infer/
infer_export:null
train_model:./inference/ch_ppocr_mobile_v2.0_det_train/best_accuracy
infer_export:tools/export_model.py -c configs/det/det_mv3_db.yml -o
infer_quant:False
inference:tools/infer/predict_det.py
--use_gpu:True|False
--enable_mkldnn:True|False
--cpu_threads:1|6
--cpu_threads:6
--rec_batch_num:1
--use_tensorrt:False|True
--precision:fp32|fp16|int8
......@@ -62,6 +62,21 @@ inference:./deploy/cpp_infer/build/ppocr det
--precision:fp32|fp16
--det_model_dir:
--image_dir:./inference/ch_det_data_50/all-sum-510/
--save_log_path:null
null:null
--benchmark:True
===========================serving_params===========================
trans_model:-m paddle_serving_client.convert
--dirname:./inference/ch_ppocr_mobile_v2.0_det_infer/
--model_filename:inference.pdmodel
--params_filename:inference.pdiparams
--serving_server:./deploy/pdserving/ppocr_det_mobile_2.0_serving/
--serving_client:./deploy/pdserving/ppocr_det_mobile_2.0_client/
serving_dir:./deploy/pdserving
web_service:web_service_det.py --config=config.yml --opt op.det.concurrency=1
op.det.local_service_conf.devices:null|0
op.det.local_service_conf.use_mkldnn:True|False
op.det.local_service_conf.thread_num:1|6
op.det.local_service_conf.use_trt:False|True
op.det.local_service_conf.precision:fp32|fp16|int8
pipline:pipeline_http_client.py --image_dir=../../doc/imgs
tests/ocr_det_server_params.txt
View file @ ac98415b
......@@ -13,7 +13,7 @@ train_infer_img_dir:./train_data/icdar2015/text_localization/ch4_test_images/
null:null
##
trainer:norm_train|pact_train
norm_train:tools/train.py -c configs/det/det_r50_vd_db.yml -o Global.pretrained_model=""
norm_train:tools/train.py -c tests/configs/det_r50_vd_db.yml -o Global.pretrained_model=""
pact_train:null
fpgm_train:null
distill_train:null
......@@ -21,13 +21,13 @@ null:null
null:null
##
===========================eval_params===========================
eval:tools/eval.py -c configs/det/det_mv3_db.yml -o
eval:tools/eval.py -c tests/configs/det_r50_vd_db.yml -o
null:null
##
===========================infer_params===========================
Global.save_inference_dir:./output/
Global.pretrained_model:
norm_export:tools/export_model.py -c configs/det/det_r50_vd_db.yml -o
norm_export:tools/export_model.py -c tests/configs/det_r50_vd_db.yml -o
quant_export:null
fpgm_export:null
distill_export:null
......
tests/ocr_ppocr_mobile_params.txt 0 → 100644
View file @ ac98415b
===========================train_params===========================
model_name:ocr_system_mobile
python:python3.7
gpu_list:null
Global.use_gpu:null
Global.auto_cast:null
Global.epoch_num:null
Global.save_model_dir:./output/
Train.loader.batch_size_per_card:null
Global.pretrained_model:null
train_model_name:null
train_infer_img_dir:null
null:null
##
trainer:
norm_train:null
pact_train:null
fpgm_train:null
distill_train:null
null:null
null:null
##
===========================eval_params===========================
eval:null
null:null
##
===========================infer_params===========================
Global.save_inference_dir:./output/
Global.pretrained_model:
norm_export:null
quant_export:null
fpgm_export:null
distill_export:null
export1:null
export2:null
##
infer_model:./inference/ch_ppocr_mobile_v2.0_det_infer/
infer_export:null
infer_quant:False
inference:tools/infer/predict_system.py
--use_gpu:True
--enable_mkldnn:True|False
--cpu_threads:1|6
--rec_batch_num:1
--use_tensorrt:False|True
--precision:fp32|fp16|int8
--det_model_dir:
--image_dir:./inference/ch_det_data_50/all-sum-510/
--save_log_path:null
--benchmark:True
--rec_model_dir:./inference/ch_ppocr_mobile_v2.0_rec_infer/
===========================cpp_infer_params===========================
use_opencv:True
infer_model:./inference/ch_ppocr_mobile_v2.0_det_infer/
infer_quant:False
inference:./deploy/cpp_infer/build/ppocr system
--use_gpu:True|False
--enable_mkldnn:True|False
--cpu_threads:1|6
--rec_batch_num:1
--use_tensorrt:False|True
--precision:fp32|fp16
--det_model_dir:
--image_dir:./inference/ch_det_data_50/all-sum-510/
--rec_model_dir:./inference/ch_ppocr_mobile_v2.0_rec_infer/
--benchmark:True
tests/ocr_ppocr_server_params.txt 0 → 100644
View file @ ac98415b
===========================train_params===========================
model_name:ocr_system_server
python:python3.7
gpu_list:null
Global.use_gpu:null
Global.auto_cast:null
Global.epoch_num:null
Global.save_model_dir:./output/
Train.loader.batch_size_per_card:null
Global.pretrained_model:null
train_model_name:null
train_infer_img_dir:null
null:null
##
trainer:
norm_train:null
pact_train:null
fpgm_train:null
distill_train:null
null:null
null:null
##
===========================eval_params===========================
eval:null
null:null
##
===========================infer_params===========================
Global.save_inference_dir:./output/
Global.pretrained_model:
norm_export:null
quant_export:null
fpgm_export:null
distill_export:null
export1:null
export2:null
##
infer_model:./inference/ch_ppocr_server_v2.0_det_infer/
infer_export:null
infer_quant:False
inference:tools/infer/predict_system.py
--use_gpu:True
--enable_mkldnn:True|False
--cpu_threads:1|6
--rec_batch_num:1
--use_tensorrt:False|True
--precision:fp32|fp16|int8
--det_model_dir:
--image_dir:./inference/ch_det_data_50/all-sum-510/
--save_log_path:null
--benchmark:True
--rec_model_dir:./inference/ch_ppocr_server_v2.0_rec_infer/
===========================cpp_infer_params===========================
use_opencv:True
infer_model:./inference/ch_ppocr_server_v2.0_det_infer/
infer_quant:False
inference:./deploy/cpp_infer/build/ppocr system
--use_gpu:True|False
--enable_mkldnn:True|False
--cpu_threads:1|6
--rec_batch_num:1
--use_tensorrt:False|True
--precision:fp32|fp16
--det_model_dir:
--image_dir:./inference/ch_det_data_50/all-sum-510/
--rec_model_dir:./inference/ch_ppocr_server_v2.0_rec_infer/
--benchmark:True
\ No newline at end of file
tests/ocr_rec_params.txt
View file @ ac98415b
===========================train_params===========================
model_name:ocr_rec
python:python3.7
gpu_list:0|2,3
gpu_list:0|0,1
Global.use_gpu:True|True
Global.auto_cast:null
Global.epoch_num:lite_train_infer=2|whole_train_infer=300
......@@ -9,7 +9,7 @@ Global.save_model_dir:./output/
Train.loader.batch_size_per_card:lite_train_infer=128|whole_train_infer=128
Global.pretrained_model:null
train_model_name:latest
train_infer_img_dir:./train_data/ic15_data/train
train_infer_img_dir:./inference/rec_inference
null:null
##
trainer:norm_train|pact_train
......@@ -41,7 +41,7 @@ inference:tools/infer/predict_rec.py
--use_gpu:True|False
--enable_mkldnn:True|False
--cpu_threads:1|6
--rec_batch_num:1
--rec_batch_num:1|6
--use_tensorrt:True|False
--precision:fp32|fp16|int8
--rec_model_dir:
......@@ -49,3 +49,33 @@ inference:tools/infer/predict_rec.py
--save_log_path:./test/output/
--benchmark:True
null:null
===========================cpp_infer_params===========================
use_opencv:True
infer_model:./inference/ch_ppocr_mobile_v2.0_rec_infer/
infer_quant:False
inference:./deploy/cpp_infer/build/ppocr rec
--use_gpu:True|False
--enable_mkldnn:True|False
--cpu_threads:1|6
--rec_batch_num:1
--use_tensorrt:False|True
--precision:fp32|fp16
--rec_model_dir:
--image_dir:./inference/rec_inference/
null:null
--benchmark:True
===========================serving_params===========================
trans_model:-m paddle_serving_client.convert
--dirname:./inference/ch_ppocr_mobile_v2.0_rec_infer/
--model_filename:inference.pdmodel
--params_filename:inference.pdiparams
--serving_server:./deploy/pdserving/ppocr_rec_mobile_2.0_serving/
--serving_client:./deploy/pdserving/ppocr_rec_mobile_2.0_client/
serving_dir:./deploy/pdserving
web_service:web_service_rec.py --config=config.yml --opt op.rec.concurrency=1
op.rec.local_service_conf.devices:null|0
op.rec.local_service_conf.use_mkldnn:True|False
op.rec.local_service_conf.thread_num:1|6
op.rec.local_service_conf.use_trt:False|True
op.rec.local_service_conf.precision:fp32|fp16|int8
pipline:pipeline_http_client.py --image_dir=../../doc/imgs_words_en
\ No newline at end of file
tests/ocr_rec_server_params.txt 0 → 100644
View file @ ac98415b
===========================train_params===========================
model_name:ocr_server_rec
python:python3.7
gpu_list:0|0,1
Global.use_gpu:True|True
Global.auto_cast:null
Global.epoch_num:lite_train_infer=2|whole_train_infer=300
Global.save_model_dir:./output/
Train.loader.batch_size_per_card:lite_train_infer=128|whole_train_infer=128
Global.pretrained_model:null
train_model_name:latest
train_infer_img_dir:./inference/rec_inference
null:null
##
trainer:norm_train|pact_train
norm_train:tools/train.py -c tests/configs/rec_icdar15_r34_train.yml -o
pact_train:deploy/slim/quantization/quant.py -c tests/configs/rec_icdar15_r34_train.yml -o
fpgm_train:null
distill_train:null
null:null
null:null
##
===========================eval_params===========================
eval:tools/eval.py -c tests/configs/rec_icdar15_r34_train.yml -o
null:null
##
===========================infer_params===========================
Global.save_inference_dir:./output/
Global.pretrained_model:
norm_export:tools/export_model.py -c tests/configs/rec_icdar15_r34_train.yml -o
quant_export:deploy/slim/quantization/export_model.py -c tests/configs/rec_icdar15_r34_train.yml -o
fpgm_export:null
distill_export:null
export1:null
export2:null
##
infer_model:./inference/ch_ppocr_server_v2.0_rec_infer/
infer_export:null
infer_quant:False
inference:tools/infer/predict_rec.py
--use_gpu:True|False
--enable_mkldnn:True|False
--cpu_threads:1|6
--rec_batch_num:1|6
--use_tensorrt:True|False
--precision:fp32|fp16|int8
--rec_model_dir:
--image_dir:./inference/rec_inference
--save_log_path:./test/output/
--benchmark:True
null:null
===========================cpp_infer_params===========================
use_opencv:True
infer_model:./inference/ch_ppocr_server_v2.0_rec_infer/
infer_quant:False
inference:./deploy/cpp_infer/build/ppocr rec
--use_gpu:True|False
--enable_mkldnn:True|False
--cpu_threads:1|6
--rec_batch_num:1
--use_tensorrt:False|True
--precision:fp32|fp16
--rec_model_dir:
--image_dir:./inference/rec_inference/
null:null
--benchmark:True
===========================serving_params===========================
trans_model:-m paddle_serving_client.convert
--dirname:./inference/ch_ppocr_server_v2.0_rec_infer/
--model_filename:inference.pdmodel
--params_filename:inference.pdiparams
--serving_server:./deploy/pdserving/ppocr_rec_server_2.0_serving/
--serving_client:./deploy/pdserving/ppocr_rec_server_2.0_client/
serving_dir:./deploy/pdserving
web_service:web_service_rec.py --config=config.yml --opt op.rec.concurrency=1
op.rec.local_service_conf.devices:null|0
op.rec.local_service_conf.use_mkldnn:True|False
op.rec.local_service_conf.thread_num:1|6
op.rec.local_service_conf.use_trt:False|True
op.rec.local_service_conf.precision:fp32|fp16|int8
pipline:pipeline_http_client.py --image_dir=../../doc/imgs_words_en
\ No newline at end of file
tests/prepare.sh
View file @ ac98415b
#!/bin/bash
FILENAME=$1
# MODE be one of ['lite_train_infer' 'whole_infer' 'whole_train_infer', 'infer', 'cpp_infer']
# MODE be one of ['lite_train_infer' 'whole_infer' 'whole_train_infer', 'infer', 'cpp_infer', 'serving_infer']
MODE=$2
dataline=$(cat ${FILENAME})
......@@ -40,11 +41,13 @@ if [ ${MODE} = "lite_train_infer" ];then
rm -rf ./train_data/ic15_data
wget -nc -P ./train_data/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/icdar2015_lite.tar
wget -nc -P ./train_data/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/ic15_data.tar # todo change to bcebos
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/rec_inference.tar
wget -nc -P ./deploy/slim/prune https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/sen.pickle
cd ./train_data/ && tar xf icdar2015_lite.tar && tar xf ic15_data.tar
ln -s ./icdar2015_lite ./icdar2015
cd ../
cd ./inference && tar xf rec_inference.tar && cd ../
elif [ ${MODE} = "whole_train_infer" ];then
wget -nc -P ./pretrain_models/ https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_large_x0_5_pretrained.pdparams
rm -rf ./train_data/icdar2015
......@@ -61,37 +64,85 @@ elif [ ${MODE} = "whole_infer" ];then
cd ./train_data/ && tar xf icdar2015_infer.tar && tar xf ic15_data.tar
ln -s ./icdar2015_infer ./icdar2015
cd ../
elif [ ${MODE} = "infer" ] || [ ${MODE} = "cpp_infer" ];then
elif [ ${MODE} = "infer" ];then
if [ ${model_name} = "ocr_det" ]; then
eval_model_name="ch_ppocr_mobile_v2.0_det_infer"
eval_model_name="ch_ppocr_mobile_v2.0_det_train"
rm -rf ./train_data/icdar2015
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/ch_det_data_50.tar
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_infer.tar
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_train.tar
cd ./inference && tar xf ${eval_model_name}.tar && tar xf ch_det_data_50.tar && cd ../
elif [ ${model_name} = "ocr_server_det" ]; then
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_infer.tar
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/ch_det_data_50.tar
cd ./inference && tar xf ch_ppocr_server_v2.0_det_infer.tar && tar xf ch_det_data_50.tar && cd ../
else
elif [ ${model_name} = "ocr_system_mobile" ]; then
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_infer.tar
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/ch_det_data_50.tar
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_infer.tar
cd ./inference && tar xf ch_ppocr_mobile_v2.0_det_infer.tar && tar xf ch_ppocr_mobile_v2.0_rec_infer.tar && tar xf ch_det_data_50.tar && cd ../
elif [ ${model_name} = "ocr_system_server" ]; then
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_infer.tar
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/ch_det_data_50.tar
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_infer.tar
cd ./inference && tar xf ch_ppocr_server_v2.0_det_infer.tar && tar xf ch_ppocr_server_v2.0_rec_infer.tar && tar xf ch_det_data_50.tar && cd ../
elif [ ${model_name} = "ocr_rec" ]; then
rm -rf ./train_data/ic15_data
eval_model_name="ch_ppocr_mobile_v2.0_rec_infer"
wget -nc -P ./train_data/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/ic15_data.tar
wget -nc -P ./inference/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/rec_inference.tar
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_infer.tar
cd ./inference && tar xf ${eval_model_name}.tar && tar xf ic15_data.tar && cd ../
cd ./inference && tar xf ${eval_model_name}.tar && tar xf rec_inference.tar && cd ../
elif [ ${model_name} = "ocr_server_rec" ]; then
rm -rf ./train_data/ic15_data
eval_model_name="ch_ppocr_server_v2.0_rec_infer"
wget -nc -P ./inference/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/rec_inference.tar
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_infer.tar
cd ./inference && tar xf ${eval_model_name}.tar && tar xf rec_inference.tar && cd ../
fi
elif [ ${MODE} = "cpp_infer" ];then
if [ ${model_name} = "ocr_det" ]; then
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/ch_det_data_50.tar
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_infer.tar
cd ./inference && tar xf ch_ppocr_mobile_v2.0_det_infer.tar && tar xf ch_det_data_50.tar && cd ../
elif [ ${model_name} = "ocr_rec" ]; then
wget -nc -P ./inference/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/rec_inference.tar
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_infer.tar
cd ./inference && tar xf ch_ppocr_mobile_v2.0_rec_infer.tar && tar xf rec_inference.tar && cd ../
elif [ ${model_name} = "ocr_system" ]; then
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_infer.tar
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/ch_det_data_50.tar
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_infer.tar
cd ./inference && tar xf ch_ppocr_mobile_v2.0_det_infer.tar && tar xf ch_ppocr_mobile_v2.0_rec_infer.tar && tar xf ch_det_data_50.tar && cd ../
fi
fi
if [ ${MODE} = "serving_infer" ];then
# prepare serving env
python_name=$(func_parser_value "${lines[2]}")
wget https://paddle-serving.bj.bcebos.com/chain/paddle_serving_server_gpu-0.0.0.post101-py3-none-any.whl
${python_name} -m pip install install paddle_serving_server_gpu-0.0.0.post101-py3-none-any.whl
${python_name} -m pip install paddle_serving_client==0.6.1
${python_name} -m pip install paddle-serving-app==0.6.3
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_infer.tar
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_infer.tar
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_infer.tar
wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_infer.tar
cd ./inference && tar xf ch_ppocr_mobile_v2.0_det_infer.tar && tar xf ch_ppocr_mobile_v2.0_rec_infer.tar && tar xf ch_ppocr_server_v2.0_rec_infer.tar && tar xf ch_ppocr_server_v2.0_det_infer.tar cd ../
fi
if [ ${MODE} = "cpp_infer" ];then
cd deploy/cpp_infer
use_opencv=$(func_parser_value "${lines[52]}")
if [ ${use_opencv} = "True" ]; then
if [ -d "opencv-3.4.7/opencv3/" ] && [ $(md5sum opencv-3.4.7.tar.gz | awk -F ' ' '{print $1}') = "faa2b5950f8bee3f03118e600c74746a" ];then
echo "################### build opencv skipped ###################"
else
echo "################### build opencv ###################"
rm -rf 3.4.7.tar.gz opencv-3.4.7/
wget https://github.com/opencv/opencv/archive/3.4.7.tar.gz
tar -xf 3.4.7.tar.gz
rm -rf opencv-3.4.7.tar.gz opencv-3.4.7/
wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/opencv-3.4.7.tar.gz
tar -xf opencv-3.4.7.tar.gz
cd opencv-3.4.7/
install_path=$(pwd)/opencv-3.4.7/opencv3
install_path=$(pwd)/opencv3
rm -rf build
mkdir build
......@@ -120,6 +171,7 @@ if [ ${MODE} = "cpp_infer" ];then
cd ../
echo "################### build opencv finished ###################"
fi
fi
echo "################### build PaddleOCR demo ####################"
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment